You could even ...
Add a stage to parsing which built a Triple to location map
(ParserProfile receives the line and column number).
The look up in that map to find the source of the triple. Imperfect in
the general case but for working on a single file it might do what you
are looking for.
Triples are safe to use as hash keys.
Andy
On 17/01/17 20:23, A. Soroka wrote:
There are several answers.
There is no reason to suppose that any given triple actually derives from a
file at all. It might have been created programmatically, or by inference, or
from SPARQL, amongst many possible other means.
You are suggesting the carriage of a really large amount of metadata all
throughout Jena's internals. The performance implications would be big, and
entirely negative.
Andy has given you a really good road to go down if what you want is more detailed
parsing metadata for, as you say, "reporting issues with the content". You can
take off that metadata and record it elsewhere or record it in RDF in various ways.
Perhaps you can tell us a little more about your use case and we can help you find a more
targeted technique for it.
---
A. Soroka
The University of Virginia Library
On Jan 17, 2017, at 3:14 PM, Grahame Grieve
<[email protected]> wrote:
hi
Yes replacing a library is not simple, but I thought I'd still make the
offer. Other advantages... no, it's just a JSON parser.
You did seem to be asking for a way to get from a triple in a graph to
the line where it was read, and that is not possible. There is no such
association.
why not? the library could provide a way, and retain the association.
Grahame
On Wed, Jan 18, 2017 at 6:58 AM, A. Soroka <[email protected]> wrote:
Replacing the JSON library in use is a considerably bigger proposition
than working with the one we now use in a different way. Are there other
advantages to using your custom code? We want to stick to well-supported
dependencies unless there is a convincing argument otherwise.
As for Turtle, I believe you can take a look at LangTurtleBase to see what
might be done. Keep in mind that there's not necessarily a precise way to
understand what line produces an error-- it might occur in the interaction
between tokens on more than one line.
---
A. Soroka
The University of Virginia Library
On Jan 17, 2017, at 2:42 PM, Grahame Grieve <
[email protected]> wrote:
well, I care about turtle and json-ld. I can contribute a json library
that preserves line numbers when the json is parsed, since the main
stream
ones don't.
Grahame
On Wed, Jan 18, 2017 at 5:38 AM, A. Soroka <[email protected]> wrote:
That will depend a bit on the language. For example, JSON parsing
doesn't
occur directly in Jena, Jena uses a library that parses from JSON to
Java
objects and then works with those objects:
org.apache.jena.riot.lang.JsonLDReader.read(InputStream, String,
ContentType, StreamRDF, Context)
In some other cases, it seems like it should be possible. Do you have a
specific language in mind?
---
A. Soroka
The University of Virginia Library
On Jan 16, 2017, at 6:48 AM, Grahame Grieve <
[email protected]> wrote:
Can the Jena parser maintain a link between the triples and the line
number
from which are sourced in the original file? This is really useful for
reporting issues with the content....
Grahame
--
-----
http://www.healthintersections.com.au / grahame@healthintersections.
com.au
/ +61 411 867 065
--
-----
http://www.healthintersections.com.au / grahame@healthintersections.
com.au
/ +61 411 867 065
--
-----
http://www.healthintersections.com.au / [email protected]
/ +61 411 867 065
